CN106023151B - Tongue object detection method under a kind of open environment - Google Patents
Tongue object detection method under a kind of open environment Download PDFInfo
- Publication number
- CN106023151B CN106023151B CN201610300494.4A CN201610300494A CN106023151B CN 106023151 B CN106023151 B CN 106023151B CN 201610300494 A CN201610300494 A CN 201610300494A CN 106023151 B CN106023151 B CN 106023151B
- Authority
- CN
- China
- Prior art keywords
- image
- value
- tongue
- obtains
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/0059—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence
- A61B5/0077—Devices for viewing the surface of the body, e.g. camera, magnifying lens
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Theoretical Computer Science (AREA)
- Pathology (AREA)
- Quality & Reliability (AREA)
- General Physics & Mathematics (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Image Analysis (AREA)
Abstract
Tongue object detection method under a kind of open environment, is related to Medical Image Processing.1) the image A acquired under open environment is inputted；2) color correction is carried out to the image A of acquisition, obtains correction image B；3) image segmentation is carried out to the correction image B that step 2) obtains；4) provincial characteristics judgement is carried out to the image C that step 3) obtains；5) textural characteristics judgement is carried out to the candidate tongue body region D that step 4) obtains.The pretreatment for carrying out color correction to image first is reduced because external light source colour temperature bring influences；Then image is split, obtains multiple connected regions；And feature judgement is carried out to each connected region, obtain candidate tongue body region；Judged judge whether candidate's tongue body region is tongue picture finally by comparison domain textural characteristics.It is finally reached the purpose for carrying out target detection to the tongue picture in image, that is, is judged either with or without tongue picture in current image, if so, tongue picture target is at which.
Description
Technical field
The present invention relates to Medical Image Processings, more particularly, to Tongue object detection method under a kind of open environment.
Background technique
Tongue picture, that is, people stretches out the tongue image presented after tongue, and the acquisition of Traditional Chinese Medicine tongue picture relies primarily on doctor's naked eyes and sees
It examines, in recent years, the development of information technology has pushed Tongue to analyze the process for objectifying, digitizing and automating.Both at home and abroad
Scholar has carried out many beneficial explorations to this, and develops some Analysis of Lingual Picture systems, achieve preferable effect (Jiang according to
My computerization TCM tongue diagnosis system [J] China combination of Chinese tradiational and Western medicine magazine, 2000,20 (2): 145-147；Cai Yihang, Liu Changjiang,
Design scheme [J] observation and control technology of the novel tongue image analysis of Shen Lan sweet-smelling grass, 2005,24 (5): 34-36.), but the bat of these systems
Take the photograph what environment was usually fixed, i.e., the shooting, collecting under the stable environment of closed, illumination, and these equipment and instruments are mostly more stupid
Weight, not portable, price costly, has certain limitation (in Liu Feng auxiliary Chinese patent medicine using system of Chinese medicine
Analysis of Lingual Picture studies the Xiamen [D]: Xiamen University, master thesis, 2007.).As smart phone, tablet computer etc. are mobile
Equipment it is universal, tongue image acquisition is carried out under open natural environment by mobile device, obtain personal health information gradually at
For a developing direction.But accompanying problem is that there are light source color temperatures, light due to acquiring tongue picture under open environment
The influence of many uncertain factors such as power, shooting angle, equipment difference, so that the image finally obtained is the same as fixed environment phase
Than, often there is larger difference in the tongue picture image of acquisition, difficulty is brought to subsequent analysis, or even due to photographer,
The tongue picture image for analysis may be not present in the image of acquisition.Therefore, when acquisition tongue picture is analyzed under open environment,
The serious forgiveness of enhancing system seems particularly significant with robustness.Before analyzing tongue picture, the inspection of tongue picture target is carried out to image
Survey helps to improve the subsequent analysis speed of system and accuracy rate, it helps doctor of traditional Chinese medicine is intuitive, it is accurate, easily observe tongue
As improving the speed of diagnosis.The purpose of tongue picture target detection is in the image of judgement acquisition with the presence or absence of suitably for subsequent point
The tongue picture of analysis, if need to be detected there are tongue picture in which position of image in the picture of acquisition.The standard of the target detection of tongue picture
True property directly affects the serious forgiveness of whole system.
Currently, the research that is unfolded for Tongue target detection under open environment and few, most connects with the method for the present invention
A kind of Tongue analysis system (Wang Bo based on mobile terminal that close technology designs for Xiamen Qiangben Science and Technology Co., Ltd
A kind of Tongue analysis system based on mobile terminal of bright: Chinese invention patent discloses, 2014200304657 [P] .2014-
12-03.), the tongue picture of mobile device acquisition detected, analyzed.Although the system realizes tongue picture detection function, but also deposit
In some shortcomings:
1, it needs to be transformed into hough space ballot to target cutting fritter to be detected, algorithm is complicated, and detection speed is slow；And
Tongue image acquisition, detection are carried out using mobile device under open environment, it is often higher to the requirement of real-time for the treatment of process；
2, it needs to describe tongue picture model with more parameter in treatment process, it is more complicated, it realizes relatively difficult；
3, the training mass data for needing supervision, needs manually to mark foreground and background, to establish detection
Model, training process difficulty are big；
4, when carrying out data training, it is desirable that sample is as more as possible, high to the dependence of data set.
Summary of the invention
It is an object of the invention to for it is existing tongue picture is acquired under open environment during there are many shortcomings,
Such as: the influence of light source color temperature, light intensity, shooting angle, many uncertain factors of equipment difference, so that the figure finally obtained
Compared with fixed environment, often there is larger difference in the tongue picture image of acquisition with picture, bring difficulty to subsequent analyze, or even due to
It the reason of photographer, may the problems such as there is no tongue picture images for analysis in the image of acquisition.The present invention provides one kind and opens
Tongue object detection method under environment is put, this method can quickly and accurately judge in the image of acquisition with the presence or absence of suitable
The tongue picture for subsequent analysis, if acquisition image in there are tongue pictures, judge tongue picture in the position of image.
The present invention the following steps are included:
1) the image A acquired under open environment is inputted；
2) color correction is carried out to the image A of acquisition, obtains correction image B；
In step 2), the image A of described pair of acquisition carries out color correction, and the specific method is as follows:
(1) tongue picture image S1 is acquired under standard light environment, calculates tri- Color Channel mean values of RGB of tongue picture image S1
Respectively with the whole mean value K of tongue picture image S1sRatio ccr、αg、αb:
Wherein, the whole mean value K of tongue picture image S1s=(Ravgs+Gavgs+Bavgs)/3；Ravgs、Gavgs、BavgsRespectively mark
The mean value of tri- Color Channels of RGB of tongue picture image S1 is acquired under quasi- light environment；The standard illumination acquisition environment can be D65
Light source, colour temperature 6500K；
(2) mean value for adjusting tri- Color Channels of RGB of image A as the following formula obtains correction image B；
K=(Ravg+Gavg+Bavg)/3
Wherein, K is the whole mean value of image A, Ravg、Gavg、BavgTri- Color Channels of RGB of respectively image A it is equal
Value；Rd、Gd、BdFor the value of tri- Color Channels of RGB of the correction each pixel of image B, Rs、Gs、BsFor each pixel of image A
Tri- Color Channels of RGB value, αr、αg、αbFor the obtained ratio in step 2) (1) part；
3) image segmentation is carried out to the correction image B that step 2) obtains, the specific method is as follows:
(1) high-ranking officers' positive image B is transformed into gray space image fB1, according to maximum variance between clusters (referring to document: OHTSU
N.A threshold selection method from gray-level histograms[J].System Man&
Cybernetics IEEE Transactions on, 1979,9 (1): 62-66.) threshold value point is carried out to gray space image fB1
It cuts, obtains segmented image B1 ', and the smooth connected domain of morphology operations is used to segmented image B1 ', obtain image B1；
In step 3) (1) part, high-ranking officers' positive image B is transformed into gray space image fB1, according between maximum kind
Variance method carries out Threshold segmentation to gray space image fB1, obtains segmented image B1 ', and use morphology to segmented image B1 '
The smooth connected domain of operation, obtaining image B1, specific step is as follows:
A) high-ranking officers' positive image B is transformed into gray space image fB1；
B) for gray space image fB1, if the value range G=[0, L-1] of the gray scale G of gray space image fB1, respectively
The probability that gray value occurs is Pi, threshold value T, threshold value T are divided into f after carrying out binaryzation to gray space image fB10And f1: f0=
[0, T], f1=[T+1, L-1], f0And f1Probability be respectivelyAnd α1=1- α0, average gray value is respectivelyWithThen f0And f1Maximum between-cluster variance are as follows: g2(T)=α0(μ0-
μ)2+α1(μ1-μ)2=α0α1(μ0-μ1)2, wherein μ=Σ iPi, threshold value T when g is maximized is found out, to gray space image
FB1 carries out Threshold segmentation, obtains the RGB value f of segmented image B1 ' pixelX, y(r, g, b):
Wherein, fB1 (x, y) indicates the value of gray space image fB1 pixel, and T is threshold value；
C) to segmented image B1 ' with the smooth connected domain of closed operation in morphology operations, according to following two formula according to
Secondary calculating obtains image B1；Pixel value g1 (x, y) in image B1 are as follows:
G1 (x, y)=erode (dilate (f1 (x, y), element))
G1 (x, y)=bitwise_not (g1 (x, y))
Wherein, f1 (x, y) is the pixel value in segmented image B1 ', and element is defined as the structural elements in morphology operations
Element；Dilate is defined as the expansive working in morphology operations；Erode is defined as the etching operation in morphology operations；
Bitwise_not is defined as the inversion operation to each pixel of image.
(2) high-ranking officers' positive image B is transformed into hsv color spatial image fB2, carries out Threshold segmentation to the channel H of image fB2,
Segmented image B2 ' is obtained, and the smooth connected domain of morphology operations is used to segmented image B2 ', obtains image B2；
In step 3) (2) part, high-ranking officers' positive image B is transformed into hsv color spatial image fB2, to image fB2
The channel H carry out Threshold segmentation, obtain segmented image B2 ', and the smooth connected domain of morphology operations is used to segmented image B2 ',
Obtaining image B2, specific step is as follows:
A) high-ranking officers' positive image B is transformed into hsv color spatial image fB2；
B) hue threshold segmentation is carried out to image fB2 using following formula, obtains the RGB value f of segmented image B2 ' pixelX, y
(r, g, b):
Wherein, hX, yIndicate the channel H pixel value in image fB2, T1And T2Indicate the threshold value of setting；
C) to segmented image B2 ' with the smooth connected domain of closed operation in morphology operations, according to following two formula according to
Secondary calculating obtains image B2；Pixel value g2 (x, y) in image B2 are as follows:
G2 (x, y)=erode (dilate (f2 (x, y), element))
G2 (x, y)=bitwise_not (g2 (x, y))
Wherein, f2 (x, y) is the pixel value in segmented image B2 ', and element is defined as the structural elements in morphology operations
Element；Dilate is defined as the expansive working in morphology operations；Erode is defined as the etching operation in morphology operations；
Bitwise_not is defined as the inversion operation to each pixel of image.
(3) using tri- colouring component variance method of RGB, (referring to document: Jiang is according in my computerization TCM tongue diagnosis system [J] China
Doctor trained in Western medicine combination magazine, 2000,20 (2): 145-147) Threshold segmentation is carried out to correction image B, segmented image B3 ' is obtained, and right
Segmented image B3 ' uses the smooth connected domain of morphology operations, obtains image B3.
It is described that Threshold segmentation is carried out to correction image B using tri- colouring component variance method of RGB in step 3) (3) part,
Segmented image B3 ' is obtained, and the smooth connected domain of morphology operations is used to segmented image B3 ', obtains the specific step of image B3
It is rapid as follows:
A) for correcting image B, it is assumed that its size is m × n, and the RGB value of each pixel carries out in high-ranking officers' positive image B
Normalization operation, value range are [0,1], calculate tri- colouring component of RGB to each pixel in correction image B using following formula
Variance gate (m, n), and correction image B is split, obtain the RGB value f of segmented image B3 ' pixelM, n(r, g, b):
Gate (m, n)=(rM, n-gM, n)+(bM, n-gM, n)×6+(rM, n+gM, n+bM, n)/3
Wherein, rM, nIndicate the value in the channel R of pixel (m, n) in correction image B, gM, nIndicate picture in correction image B
The value in the channel G of vegetarian refreshments (m, n), bM, nIndicate the value of the channel B of pixel (m, n) in correction image B；
B) segmented image B3 ' is calculated according to following formula, is obtained with the smooth connected domain of closed operation in morphology operations
To image B3；Pixel value g3 (x, y) in image B3 are as follows:
G3 (x, y)=erode (dilate (f3 (x, y), element))
Wherein, f3 (x, y) is the pixel value in segmented image B3 ', and element is defined as the structural elements in morphology operations
Element；Dilate is defined as the expansive working in morphology operations；Erode is defined as the etching operation in morphology operations.
(4) logic "and" operation is carried out to tri- image B1, image B2, image B3 images, obtains image C.
4) provincial characteristics judgement is carried out to the image C that step 3) obtains, the specific method is as follows:
(1) for each of image C connected domain, the convex closure (S of the connected domain is calculatedi)；
(2) each convex closure (S is calculatedi) areaWith the area area of image CCRatio:Delete connected domain of the area ratio less than 0.02；
(4) each connected domain convex closure (S is calculatedi) mass center (Ci) and image C center (C0) Euclidean distance:
Wherein, Cix、CiyRespectively indicate convex closure (Si) mass center abscissa and ordinate, C0xAnd C0yIt respectively indicates in image C
The abscissa and ordinate of the heart；
(4) each connected domain convex closure (S is calculatedi) minimum circumscribed rectangle length-width ratio scale=w/h；
Wherein w indicates that the width of minimum circumscribed rectangle, h indicate the length of minimum circumscribed rectangle；
(5) each connected domain convex closure (S is calculatedi) 7 Hu not bending moment mi(i ∈ [1,7] is (referring to document: Hu M.Visual
Pattern Recognition by Moment Invariants[J].Information Theory Ire
Transactions on,1962,8(2):179-187.)；
Bending moment is not defined as follows the Hu:
The image f (x, y) for being M × N for size, two dimension (p+q) rank square of f (x, y) is defined as:
(p+q) rank central moment accordingly is defined as:
Wherein
By ηpqThe normalization central moment of expression is defined as:
Wherein γ=(p+q)/2+1
Hu is constructed with the linear combination of normalization central moment has translation, flexible, invariable rotary 7 Hu not bending moment
(6) normal shape tongue picture image (S is opened in artificial selection one0), calculate tongue picture image (S0) 7 Hu not bending moment Mi(i
∈ [1,7]), calculate matching degreeThe normal shape tongue picture image (S0) can be cured for Chinese medicine
The raw rule of thumb normal shape tongue picture image with relevant professional knowledge selection；
(7) each connected domain convex closure (Si is successively calculated according to following three formula)Similarity score, select similarity
The maximum value of score, the corresponding connected domain convex closure (S of the maximum value of similarity scorei) it is the candidate tongue body region D selected:
Wherein, area is in the obtained ratio in step 4) (2) part；Scale is acquired by step 4) (4) part
Length-width ratio；Dis is the obtained Euclidean distance in step 4) (3) part；Match is that step 4) (6) part is obtained
Matching degree；
5) textural characteristics judgement is carried out to the candidate tongue body region D that step 4) obtains, the specific method is as follows:
(1) horizontal, vertical, three kinds of gray level co-occurrence matrixes of angular shift amount the joints of candidate tongue body region D are calculated separately
Probability density distribution；
(2) the joint probability density distribution for setting gray level co-occurrence matrixes is denoted as [Pmn] L × L, wherein L is gray scale value range, m
=[0, L-1], n=[0, L-1]；According to joint probability density distribution be calculated 6 textural characteristics (referring to document:
HARALICK R M,SHANMUGAM K,DINSTEIN I.Textural features for image
classification[J].Systems Man&Cybernetics IEEE Transactions on,2010,smc-3(6):
610-621.), 6 textural characteristics include that angular second moment, contrast, entropy, inverse difference moment, intermediate value are related to gray scale；
Wherein,
(3) three kinds of gray level co-occurrence matrixes that step 5) (1) is partially obtained joint probability density distribution respectively according to
The method of step 5) (2) part calculates 6 textural characteristics, and 18 texture characteristic amount f of candidate tongue body region D are obtainedi(i
∈[0.17])；
(4) the tongue picture image acquired under several width standard light environments of artificial selection, under each width standard light environment
The tongue picture image of acquisition is partitioned into tongue body region SD therein by hand, several tongue body regions SD is obtained, to each tongue body area
Domain SD calculates separately its horizontal, vertical, three kinds of gray level co-occurrence matrixes of angular shift amount joint probability density distribution；To these three
Joint probability density distribution calculates textural characteristics according to the method for step 5) (2) part respectively, each tongue body region SD is total
Obtain 18 texture characteristic amounts；To 18 texture characteristic amounts that these tongue body regions SD is calculated, each texture characteristic amount is sought
Average value Fi(i ∈ [0,17])；
(5) texture characteristic amount f step 5) (3) partially acquired is calculatedi(i ∈ [0,17]) and step 5) (4) part
The average value F of the texture characteristic amount acquirediThe texture similarity of (i ∈ [0,17]), is calculated using following equation, if result E is small
In the threshold value T3 of setting, then it is judged as tongue picture, there is no the tongue pictures for analysis in the image A otherwise acquired:
The present invention is directed to the characteristics of Tongue image under open environment progress tongue picture target detection, carries out first to image
The pretreatment of color correction is reduced because external light source colour temperature bring influences；Then image is split, obtains multiple connections
Region；And feature judgement is carried out to each connected region, obtain candidate tongue body region；It is carried out finally by comparison domain textural characteristics
Judgement, judges whether candidate's tongue body region is tongue picture.The present invention, which is finally reached, carries out target detection to the tongue picture in image
Purpose judges either with or without tongue picture in current image, if so, tongue picture target is at which.
Color correction process uses improved gray world algorithm, sets for the distinctive color characteristic of tongue picture image corresponding
Parameter value；Tongue picture partitioning portion is combined using maximum between-cluster variance, hue threshold segmentation and the segmentation of tri- colouring component difference of RGB
Method carry out image segmentation；Provincial characteristics judgment part is judged by comparing the shape feature in each region, obtains tongue body
Region；The textural characteristics judgement of tongue picture, which is then characterized using gray level co-occurrence matrixes, to be detected, and is carried out to the tongue body region of acquisition
Judgement compares, and is finally reached the purpose of tongue picture target detection.
Compared with the existing technology, the invention has the benefit that
1, the present invention according to the improved gray world algorithm of tongue picture characteristic use to image carry out color correction, reduce because
Acquiring environmental colors colour cast bring influences.
2, the present invention divides the side combined using maximum between-cluster variance, hue threshold segmentation and tri- colouring component difference of RGB
Method carries out image segmentation；Provincial characteristics judgement is carried out according to tongue picture shape, position；It is detected using textural characteristics；These sides
Method is handled both for the tongue picture feature under open environment, is analyzed, these methods complement each other, and being used in combination with can
It more accurately identifies tongue picture target, has the advantages that recognition accuracy is high.
3, the method operand that the present invention uses is small, program is realized simply, does not need to train mass data, algorithm complexity
Low, processing speed is fast, not high to the degree of dependence of data set.
The present invention is directed to the characteristics of tongue picture image under open environment point multiple steps and handles, analyzes, and reaches tongue picture mesh
Mark the purpose of detection.Image segmentation of the invention, provincial characteristics judgement, textural characteristics judge several steps functionally mutually auxiliary phase
Carry out coarse sizing using the color characteristic of image at, image segmentation, provincial characteristics judgement using the shape of image, space characteristics into
Row fine screening, textural characteristics judgement finally judged using the textural characteristics of image, by conjunction with image different characteristic into
Row comparison, the accuracy rate of identification is higher, and arithmetic speed is fast, realizes simply, achieves new technical effect.Relative to immediate
The prior art has apparent distinguishing characteristics, and has the advantages that obvious: method of the invention does not need hough space throwing
Ticket, algorithm complexity is low, and operand is small, does not also need the training mass data for carrying out having supervision, therefore realization is relatively simple, right
The degree of dependence of data set is not high, accuracy rate with it is more superior in processing speed, have more practicability.
It is obtaining through the invention as a result, being conducive to improve serious forgiveness, robustness and the system of subsequent lingual diagnosis analysis system
Analysis speed, be also convenient for doctor of traditional Chinese medicine it is intuitive, it is accurate, easily observe tongue picture, improve the speed of diagnosis.Of the invention is direct
Purpose is not the diagnostic result for obtaining disease or health status.Even if through the invention obtain as a result, can only obtain figure
The conclusion that whether there is the position of tongue picture and tongue picture as in, can not immediately arrive at diagnostic result, need subsequent lingual diagnosis point
The further analysis and judgement of analysis system or doctor of traditional Chinese medicine.Therefore method disclosed by the invention, direct purpose are not diagnosis,
The present invention is not belonging to the diagnostic method of disease.
Detailed description of the invention
Fig. 1 is the image acquired under open environment；
Fig. 2 is the result of step 3) image segmentation；
Fig. 3 is that step 4) screens obtained candidate tongue body region.
Specific embodiment
Technical solution of the present invention is described in further detail with reference to the accompanying drawings and detailed description:
The embodiment of the present invention includes following steps:
1) the image A acquired under open environment is inputted, as shown in Figure 1；
2) color correction is carried out to the image A of acquisition, obtains correction image B；
In step 2), the image A of described pair of acquisition carries out color correction, and the specific method is as follows:
(1) tongue picture image S1 is acquired under standard light environment, calculates tri- Color Channel mean values of RGB of tongue picture image S1
Respectively with the whole mean value K of tongue picture image S1sRatio ccr、αg、αb:
Wherein, the whole mean value K of tongue picture image S1s=(Ravgs+Gavgs+Bavgs)/3；Ravgs、Gavgs、BavgsRespectively mark
The mean value of tri- Color Channels of RGB of tongue picture image S1 is acquired under quasi- light environment；The standard illumination acquisition environment can be D65
Light source, colour temperature 6500K；
In the present embodiment, the value of parameters is as follows: αr=1.09, αg=0.95, αb=0.94.
(2) mean value for adjusting tri- Color Channels of RGB of image A as the following formula obtains correction image B；
K=(Ravg+Gavg+Bavg)/3
Wherein, K is the whole mean value of image A, Ravg、Gavg、BavgTri- Color Channels of RGB of respectively image A it is equal
Value；Rd、Gd、BdFor the value of tri- Color Channels of RGB of the correction each pixel of image B, Rs、Gs、BsFor each pixel of image A
Tri- Color Channels of RGB value, αr、αg、αbFor the obtained ratio in step 2) (1) part；
3) image segmentation is carried out to the correction image B that step 2) obtains, the specific method is as follows:
(1) high-ranking officers' positive image B is transformed into gray space image fB1, according to maximum variance between clusters to gray space image
FB1 carries out Threshold segmentation, obtains segmented image B1 ', and use the smooth connected domain of morphology operations to segmented image B1 ', obtains
Image B1；
In step 3) (1) part, high-ranking officers' positive image B is transformed into gray space image fB1, according between maximum kind
Variance method carries out Threshold segmentation to gray space image fB1, obtains segmented image B1 ', and use morphology to segmented image B1 '
The smooth connected domain of operation, obtaining image B1, specific step is as follows:
A) high-ranking officers' positive image B is transformed into gray space image fB1；
B) for gray space image fB1, if the value range G=[0, L-1] of the gray scale G of gray space image fB1, respectively
The probability that gray value occurs is Pi, threshold value T, threshold value T are divided into f after carrying out binaryzation to gray space image fB10And f1: f0=
[0, T], f1=[T+1, L-1], f0And f1Probability be respectivelyAnd α1=1- α0, average gray value is respectivelyWithThen f0And f1Maximum between-cluster variance are as follows: g2(T)=α0(μ0-
μ)2+α1(μ1-μ)2=α0α1(μ0-μ1)2, whereinThreshold value T when g is maximized is found out, to gray space figure
As fB1 progress Threshold segmentation, the RGB value f of segmented image B1 ' pixel is obtainedX, y(r, g, b):
Wherein, fB1 (x, y) indicates the value of gray space image fB1 pixel, and T is threshold value；
In the present embodiment, T=90.
C) to segmented image B1 ' with the smooth connected domain of closed operation in morphology operations, according to following two formula according to
Secondary calculating obtains image B1；Pixel value g1 (x, y) in image B1 are as follows:
G1 (x, y)=erode (dilate (f1 (x, y), element))
G1 (x, y)=bitwise_not (g1 (x, y))
Wherein, f1 (x, y) is the pixel value in segmented image B1 ', and element is defined as the structural elements in morphology operations
Element；Dilate is defined as the expansive working in morphology operations；Erode is defined as the etching operation in morphology operations；
Bitwise_not is defined as the inversion operation to each pixel of image.
In the present embodiment, element is defined as the oval structure of [11 × 11].
(2) high-ranking officers' positive image B is transformed into hsv color spatial image fB2, carries out Threshold segmentation to the channel H of image fB2,
Segmented image B2 ' is obtained, and the smooth connected domain of morphology operations is used to segmented image B2 ', obtains image B2；
In step 3) (2) part, high-ranking officers' positive image B is transformed into hsv color spatial image fB2, to image fB2
The channel H carry out Threshold segmentation, obtain segmented image B2 ', and the smooth connected domain of morphology operations is used to segmented image B2 ',
Obtaining image B2, specific step is as follows:
A) high-ranking officers' positive image B is transformed into hsv color spatial image fB2；
B) hue threshold segmentation is carried out to image fB2 using following formula, obtains the RGB value f of segmented image B2 ' pixelX, y
(r, g, b):
Wherein, hX, yIndicate the channel H pixel value in image fB2, T1And T2Indicate the threshold value of setting；
In the present embodiment, T1=7, T2=29.
C) to segmented image B2 ' with the smooth connected domain of closed operation in morphology operations, according to following two formula according to
Secondary calculating obtains image B2；Pixel value g2 (x, y) in image B2 are as follows:
G2 (x, y)=erode (dilate (f2 (x, y), element))
G2 (x, y)=bitwise_not (g2 (x, y))
Wherein, f2 (x, y) is the pixel value in segmented image B2 ', and element is defined as the structural elements in morphology operations
Element；Dilate is defined as the expansive working in morphology operations；Erode is defined as the etching operation in morphology operations；
Bitwise_not is defined as the inversion operation to each pixel of image.
In the present embodiment, element is defined as the oval structure of [11 × 11]
(3) Threshold segmentation is carried out to correction image B using tri- colouring component variance method of RGB, obtains segmented image B3 ', and right
Segmented image B3 ' uses the smooth connected domain of morphology operations, obtains image B3.
It is described that Threshold segmentation is carried out to correction image B using tri- colouring component variance method of RGB in step 3) (3) part,
Segmented image B3 ' is obtained, and the smooth connected domain of morphology operations is used to segmented image B3 ', obtains the specific step of image B3
It is rapid as follows:
A) for correcting image B, it is assumed that its size is m × n, and the RGB value of each pixel carries out in high-ranking officers' positive image B
Normalization operation, value range are [0,1], calculate tri- colouring component of RGB to each pixel in correction image B using following formula
Variance gate (m, n), and correction image B is split, obtain the RGB value f of segmented image B3 ' pixelM, n(r, g, b):
Gate (m, n)=(rM, n-gM, n)+(bM, n-gM, n)x6+(rM, n+gM, n+bM, n)/3
Wherein, rM, nIndicate the value in the channel R of pixel (m, n) in correction image B, gM, nIndicate picture in correction image B
The value in the channel G of vegetarian refreshments (m, n), bM, nIndicate the value of the channel B of pixel (m, n) in correction image B；
B) segmented image B3 ' is calculated according to following formula, is obtained with the smooth connected domain of closed operation in morphology operations
To image B3；Pixel value g3 (x, y) in image B3 are as follows:
G3 (x, y)=erode (dilate (f3 (x, y), element))
Wherein, f3 (x, y) is the pixel value in segmented image B3 ', and element is defined as the structural elements in morphology operations
Element；Dilate is defined as the expansive working in morphology operations；Erode is defined as the etching operation in morphology operations.
In the present embodiment, element is defined as the oval structure of [11 × 11]
(4) logic "and" operation is carried out to tri- image B1, image B2, image B3 images, image C is obtained, such as Fig. 2 institute
Show.
4) provincial characteristics judgement is carried out to the image C that step 3) obtains, the specific method is as follows:
(1) for each of image C connected domain, the convex closure (S of the connected domain is calculatedi)；
(2) each convex closure (S is calculatedi) areaWith the area area of image CCRatio:Delete connected domain of the area ratio less than 0.02；
(5) each connected domain convex closure (S is calculatedi) mass center (Ci) and image C center (C0) Euclidean distance:
Wherein, Cix、CiyRespectively indicate convex closure (Si) mass center abscissa and ordinate, C0xAnd C0yIt respectively indicates in image C
The abscissa and ordinate of the heart；
(4) each connected domain convex closure (S is calculatedi) minimum circumscribed rectangle length-width ratio scale=w/h；
Wherein w indicates that the width of minimum circumscribed rectangle, h indicate the length of minimum circumscribed rectangle；
(5) each connected domain convex closure (S is calculatedi) 7 Hu not bending moment mi(i ∈ [1,7])；
Bending moment is not defined as follows the Hu:
The image f (x, y) for being M × N for size, two dimension (p+q) rank square of f (x, y) is defined as:
(p+q) rank central moment accordingly is defined as:
Wherein
By ηpqThe normalization central moment of expression is defined as:
Wherein γ=(p+q)/2+1
Hu is constructed with the linear combination of normalization central moment has translation, flexible, invariable rotary 7 Hu not bending moment:
(6) normal shape tongue picture image (S is opened in artificial selection one0), calculate tongue picture image (S0) 7 Hu not bending moment Mi(i
∈ [1,7]), calculate matching degreeThe normal shape tongue picture image (S0) can be cured for Chinese medicine
The raw rule of thumb normal shape tongue picture image with relevant professional knowledge selection；
In the present embodiment, MiThe value of (i ∈ [1,7]) be respectively as follows: [0.167538,0.00155933,0.00032405,
6.71027×10-6,-3.12169×10-10,-2.63866×10-7,-2.15107×10-11]
(7) each connected domain convex closure (S is successively calculated according to following three formulai) similarity score, select similarity
The maximum value of score, the corresponding connected domain convex closure (S of the maximum value of similarity scorei) it is the candidate tongue body region D selected,
It is as shown in Figure 3:
Wherein, area is the obtained ratio in step 4) (2) part；Scale is that step 4) (4) part is obtained
Length-width ratio；Dis is the obtained Euclidean distance in step 4) (3) part；Match is obtained of step 4) (6) part
With degree；
5) textural characteristics judgement is carried out to the candidate tongue body region D that step 4) obtains, the specific method is as follows:
(1) horizontal, vertical, three kinds of gray level co-occurrence matrixes of angular shift amount the joints of candidate tongue body region D are calculated separately
Probability density distribution；
(2) the joint probability density distribution for setting gray level co-occurrence matrixes is denoted as [Pmn]L×L, wherein L is gray scale value range, m
=[0, L-1], n=[0, L-1]；6 textural characteristics, 6 textural characteristics are calculated according to joint probability density distribution
It is related to gray scale including angular second moment, contrast, entropy, inverse difference moment, intermediate value；
Wherein,
(3) three kinds of gray level co-occurrence matrixes that step 5) (1) is partially obtained joint probability density distribution respectively according to
The method of step 5) (2) part calculates 6 textural characteristics, and 18 texture characteristic amount f of candidate tongue body region D are obtainedi(i
∈ [0,17]；
(4) the tongue picture image acquired under several width standard light environments of artificial selection, under each width standard light environment
The tongue picture image of acquisition is partitioned into tongue body region SD therein by hand, obtains several tongue body regions SD；To each tongue body area
Domain SD calculates separately its horizontal, vertical, three kinds of gray level co-occurrence matrixes of angular shift amount joint probability density distribution；To these three
Joint probability density distribution calculates textural characteristics according to the method for step 5) (2) part respectively, each tongue body region SD is total
Obtain 18 texture characteristic amounts；To 18 texture characteristic amounts that these tongue body regions SD is calculated, each texture characteristic amount is sought
Average value Fi(i ∈ [0,17]；
(5) texture characteristic amount f step 5) (3) partially acquired is calculatedi(i ∈ [0,17]) and step 5) (4) part
The texture characteristic amount average value F acquirediThe texture similarity of (i ∈ [0,17]), is calculated using following equation, if result E is less than
The threshold value T3 of setting, then be judged as tongue picture, and there is no the tongue pictures for analysis in the image A otherwise acquired:
In the present embodiment, standard tongue picture feature vector average value Fi(value of i ∈ [0,17] is respectively as follows:
[0.00280036,59.4779,0.428998,2.9444,136.397,939.903,0.00197886,
68.4362,0.348336,3.0673,136.616,1336.41,0.00292386,38.6503,0.442026,2.9221,
136.406,901.906], the value of threshold value T3 is 4.7.
Claims (8)
1. Tongue object detection method under a kind of open environment, it is characterised in that the following steps are included:
1) the image A acquired under open environment is inputted；
2) color correction is carried out to the image A of acquisition, obtains correction image B；
The image A of described pair of acquisition carries out color correction, and the specific method is as follows:
(1) tongue picture image S1 is acquired under standard light environment, calculates tri- Color Channel mean value difference of RGB of tongue picture image S1
With the whole mean value K of tongue picture image S1sRatio ccr、αg、αb:
Wherein, the whole mean value K of tongue picture image S1s=(Ravgs+Gavgs+Bavgs)/3；Ravgs、Gavgs、BavgsRespectively standard illumination
The mean value of tri- Color Channels of RGB of tongue picture image S1 is acquired under environment；The standard illumination acquisition environment is D65 light source, color
Temperature is 6500K；
(2) mean value for adjusting tri- Color Channels of RGB of image A as the following formula obtains correction image B；
K=(Ravg+Gavg+Bavg)/3
Wherein, K is the whole mean value of image A, Ravg、Gavg、BavgThe mean value of tri- Color Channels of RGB of respectively image A；Rd、
Gd、BdFor the value of tri- Color Channels of RGB of the correction each pixel of image B, Rs、Gs、BsFor the RGB of each pixel of image A
The value of three Color Channels, αr、αg、αbFor the obtained ratio in step 2) (1) part；
3) image segmentation is carried out to the correction image B that step 2) obtains, obtains image C；
4) provincial characteristics judgement is carried out to the image C that step 3) obtains, obtains candidate tongue body region D；
5) textural characteristics judgement is carried out to the candidate tongue body region D that step 4) obtains.
2. Tongue object detection method under a kind of open environment as described in claim 1, it is characterised in that in step 3),
The correction image B obtained to step 2) carries out image segmentation, and the specific method is as follows:
(1) high-ranking officers' positive image B is transformed into gray space image fB1, according to maximum variance between clusters to gray space image fB1 into
Row threshold division obtains segmented image B1 ', and uses the smooth connected domain of morphology operations to segmented image B1 ', obtains image
B1；
(2) high-ranking officers' positive image B is transformed into hsv color spatial image fB2, carries out Threshold segmentation to the channel H of image fB2, obtains
Segmented image B2 ', and the smooth connected domain of morphology operations is used to segmented image B2 ', obtain image B2；
(3) Threshold segmentation is carried out to correction image B using tri- colouring component variance method of RGB, obtains segmented image B3 ', and to segmentation
Image B3 ' uses the smooth connected domain of morphology operations, obtains image B3；
(4) logic "and" operation is carried out to tri- image B1, image B2, image B3 images, obtains image C.
3. Tongue object detection method under a kind of open environment as claimed in claim 2, it is characterised in that in step 3) the
(1) in part, high-ranking officers' positive image B is transformed into gray space image fB1, according to maximum variance between clusters to gray space figure
As fB1 progress Threshold segmentation, segmented image B1 ' is obtained, and the smooth connected domain of morphology operations is used to segmented image B1 ', is obtained
To image B1, specific step is as follows:
A) high-ranking officers' positive image B is transformed into gray space image fB1；
B) for gray space image fB1, if the value range G=[0, L-1] of the gray scale G of gray space image fB1, each gray scale
The probability that value occurs is Pi, threshold value T, threshold value T are divided into f after carrying out binaryzation to gray space image fB10And f1: f0=[0,
T],f1=[T+1, L-1], f0And f1Probability be respectivelyAnd α1=1- α0, average gray value is respectivelyWithThen f0And f1Maximum between-cluster variance are as follows: g2(T)=α0(μ0-
μ)2+α1(μ1-μ)2=α0α1(μ0-μ1)2, wherein μ=∑ iPi, threshold value T when g is maximized is found out, to gray space image
FB1 carries out Threshold segmentation, obtains the RGB value f of segmented image B1 ' pixelX, y(r, g, b):
Wherein, fB1 (x, y) indicates the value of gray space image fB1 pixel, and T is threshold value；
C) segmented image B1 ' is successively counted with the smooth connected domain of closed operation in morphology operations according to following two formula
It calculates, obtains image B1；Pixel value g1 (x, y) in image B1 are as follows:
G1 (x, y)=erode (dilate (f1 (x, y), element))
G1 (x, y)=bitwise_not (g1 (x, y))
Wherein, f1 (x, y) is the pixel value in segmented image B1 ', and element is defined as the structural element in morphology operations；
Dilate is defined as the expansive working in morphology operations；Erode is defined as the etching operation in morphology operations；bitwise_
Not is defined as the inversion operation to each pixel of image.
4. Tongue object detection method under a kind of open environment as claimed in claim 2, it is characterised in that in step 3) the
(2) in part, high-ranking officers' positive image B is transformed into hsv color spatial image fB2, carries out threshold value point to the channel H of image fB2
It cuts, obtains segmented image B2 ', and the smooth connected domain of morphology operations is used to segmented image B2 ', obtain the specific step of image B2
It is rapid as follows:
A) high-ranking officers' positive image B is transformed into hsv color spatial image fB2；
B) hue threshold segmentation is carried out to image fB2 using following formula, obtains the RGB value f of segmented image B2 ' pixelX, y(r,
G, b):
Wherein, hX, yIndicate the channel H pixel value in image fB2, T1And T2Indicate the threshold value of setting；
C) segmented image B2 ' is successively counted with the smooth connected domain of closed operation in morphology operations according to following two formula
It calculates, obtains image B2；Pixel value g2 (x, y) in image B2 are as follows:
G2 (x, y)=erode (dilate (f2 (x, y), element))
G2 (x, y)=bitwise_not (g2 (x, y))
Wherein, f2 (x, y) is the pixel value in segmented image B2 ', and element is defined as the structural element in morphology operations；
Dilate is defined as the expansive working in morphology operations；Erode is defined as the etching operation in morphology operations；bitwise_
Not is defined as the inversion operation to each pixel of image.
5. Tongue object detection method under a kind of open environment as claimed in claim 2, it is characterised in that in step 3) the
(3) described that Threshold segmentation is carried out to correction image B using tri- colouring component variance method of RGB in part, segmented image B3 ' is obtained, and
The smooth connected domain of morphology operations is used to segmented image B3 ', obtaining image B3, specific step is as follows:
A) for correcting image B, it is assumed that its size is m × n, and the RGB value of each pixel carries out normalizing in high-ranking officers' positive image B
Change operation, value range is [0,1], calculates tri- colouring component variance of RGB to each pixel in correction image B using following formula
Gate (m, n), and correction image B is split, obtain the RGB value f of segmented image B3 ' pixelM, n(r, g, b):
Gate (m, n)=(rM, n-gM, n)+(bM, n-gM, n)×6+(rM, n+gM, n+bM, n)/3
Wherein, rM, nIndicate the value in the channel R of pixel (m, n) in correction image B, gM, nIndicate pixel in correction image B
The value in the channel G of (m, n), bM, nIndicate the value of the channel B of pixel (m, n) in correction image B；
B) segmented image B3 ' is calculated according to following formula with the smooth connected domain of closed operation in morphology operations, obtains figure
As B3；Pixel value g3 (x, y) in image B3 are as follows:
G3 (x, y)=erode (dilate (f3 (x, y), element))
Wherein, f3 (x, y) is the pixel value in segmented image B3 ', and element is defined as the structural element in morphology operations；
Dilate is defined as the expansive working in morphology operations；Erode is defined as the etching operation in morphology operations.
6. Tongue object detection method under a kind of open environment as described in claim 1, it is characterised in that in step 4),
The image C obtained to step 3) carries out provincial characteristics judgement, and the specific method is as follows:
(1) for each of image C connected domain, the convex closure S of the connected domain is calculatedi；
(2) each convex closure S is calculatediAreaWith the area area of image CCRatio:Delete connected domain of the area ratio less than 0.02；
(3) each connected domain convex closure S is calculatediMass center CiWith image C center C0Euclidean distance:
Wherein, Cix、CiyRespectively indicate convex closure SiThe abscissa and ordinate of mass center, C0xAnd C0yRespectively indicate the cross at the center image C
Coordinate and ordinate；
(4) each connected domain convex closure S is calculatediThe length-width ratio scale=w/h of minimum circumscribed rectangle；
Wherein w indicates that the width of minimum circumscribed rectangle, h indicate the length of minimum circumscribed rectangle；
(5) each connected domain convex closure S is calculatedi7 Hu not bending moment mi(i ∈ [1,7])；
Bending moment is not defined as follows the Hu:
The image f (x, y) for being M × N for size, two dimension (p+q) rank square of f (x, y) is defined as:
(p+q) rank central moment accordingly is defined as:
Wherein
By ηpqThe normalization central moment of expression is defined as:
Wherein γ=(p+q)/2+1
Hu is constructed with the linear combination of normalization central moment has translation, flexible, invariable rotary 7 Hu not bending moment
(6) normal shape tongue picture image S is opened in artificial selection one0, calculate tongue picture image S07 Hu not bending moment Mi(i ∈ [1,7]),
Calculate matching degree
(7) each connected domain convex closure S is successively calculated according to following three formulaiSimilarity score, select similarity score
Maximum value, the corresponding connected domain convex closure S of the maximum value of similarity scoreiFor the candidate tongue body region D selected:
Wherein, area is in the obtained ratio in step 4) (2) part；Scale is the obtained length in step 4) (4) part
Wide ratio；Dis is the obtained Euclidean distance in step 4) (3) part；Match is the obtained matching in step 4) (6) part
Degree.
7. Tongue object detection method under a kind of open environment as claimed in claim 6, it is characterised in that in step 4) the
(6) in part, the normal shape tongue picture image S0The normal shape rule of thumb selected with relevant professional knowledge for doctor of traditional Chinese medicine
Shape tongue picture image.
8. Tongue object detection method under a kind of open environment as described in claim 1, it is characterised in that in step 5),
The candidate tongue body region D obtained to step 4) carries out textural characteristics judgement, and the specific method is as follows:
(1) horizontal, vertical, three kinds of gray level co-occurrence matrixes of angular shift amount the joint probabilities of candidate tongue body region D are calculated separately
Density Distribution；
(2) the joint probability density distribution for setting gray level co-occurrence matrixes is denoted as [Pmn]L×L, wherein L be gray scale value range, m=[0,
L-1], n=[0, L-1]；6 textural characteristics are calculated according to joint probability density distribution, 6 textural characteristics include angle
Second moment, contrast, entropy, inverse difference moment, intermediate value are related to gray scale；
Wherein,
(3) joint probability density of the three kinds of gray level co-occurrence matrixes partially obtained to step 5) (1) is distributed respectively according to step
5) method of (2) part calculates 6 textural characteristics, and 18 texture characteristic amount f of candidate tongue body region D are obtainedi(i ∈ [0,
17])；
(4) the tongue picture image acquired under several width standard light environments of artificial selection, to being acquired under each width standard light environment
Tongue picture image, be partitioned into tongue body region SD therein by hand, obtain several tongue body regions SD, to each tongue body region SD
Calculate separately its horizontal, vertical, three kinds of gray level co-occurrence matrixes of angular shift amount joint probability density distribution；To these three joints
Probability density distribution calculates textural characteristics according to the method for step 5) (2) part respectively, each tongue body region SD is obtained
18 texture characteristic amounts；To 18 texture characteristic amounts that these tongue body regions SD is calculated, the flat of each texture characteristic amount is sought
Mean value Fi (i ∈ [0,17])；
(5) texture characteristic amount f step 5) (3) partially acquired is calculatediWhat (i ∈ [0,17]) and step 5) (4) partially acquired
The texture similarity of the average value Fi (i ∈ [0,17]) of texture characteristic amount, is calculated using following equation, if result E is less than setting
Threshold value T3, then be judged as tongue picture, in the image A otherwise acquired there is no for analysis tongue picture:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610300494.4A CN106023151B (en) | 2016-05-09 | 2016-05-09 | Tongue object detection method under a kind of open environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610300494.4A CN106023151B (en) | 2016-05-09 | 2016-05-09 | Tongue object detection method under a kind of open environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106023151A CN106023151A (en) | 2016-10-12 |
| CN106023151B true CN106023151B (en) | 2019-02-01 |
Family
ID=57099693
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610300494.4A Active CN106023151B (en) | 2016-05-09 | 2016-05-09 | Tongue object detection method under a kind of open environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106023151B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845313A (en) * | 2016-12-28 | 2017-06-13 | 广州智慧城市发展研究院 | A kind of binary processing method of Quick Response Code |
| CN107194937B (en) * | 2017-05-27 | 2020-04-24 | 厦门大学 | Traditional Chinese medicine tongue picture image segmentation method in open environment |
| CN108303420A (en) * | 2017-12-30 | 2018-07-20 | 上饶市中科院云计算中心大数据研究院 | A kind of domestic type sperm quality detection method based on big data and mobile Internet |
| CN108388874B (en) * | 2018-03-05 | 2020-03-31 | 厦门大学 | Prawn morphological parameter automatic measurement method based on image recognition and cascade classifier |
| CN108830874A (en) * | 2018-04-19 | 2018-11-16 | 麦克奥迪（厦门）医疗诊断系统有限公司 | A kind of number pathology full slice Image blank region automatic division method |
| CN108629780B (en) * | 2018-04-23 | 2020-01-03 | 闽江学院 | Tongue image segmentation method based on color decomposition and threshold technology |
| CN110866932A (en) * | 2018-08-20 | 2020-03-06 | 深圳市前海安测信息技术有限公司 | Multi-channel tongue edge detection device and method and storage medium |
| CN109636864A (en) * | 2018-12-19 | 2019-04-16 | 新绎健康科技有限公司 | A kind of tongue dividing method and system based on color correction Yu depth convolutional neural networks |
| CN112489053B (en) * | 2020-11-26 | 2021-07-13 | 深圳市艾合芯科智慧医疗有限公司 | Tongue image segmentation method and device and storage medium |
| CN113160203B (en) * | 2021-04-30 | 2023-06-23 | 湄洲湾职业技术学院 | Artificial intelligent tongue picture and tongue fur color recognition system |
| CN114596621B (en) * | 2022-05-10 | 2022-07-29 | 慧医谷中医药科技(天津)股份有限公司 | Tongue picture data processing method and system based on machine vision |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104021566A (en) * | 2014-06-24 | 2014-09-03 | 天津大学 | GrabCut algorithm-based automatic segmentation method of tongue diagnosis image |
| CN104933723A (en) * | 2015-07-21 | 2015-09-23 | 闽江学院 | Tongue image segmentation method based on sparse representation |
-
2016
- 2016-05-09 CN CN201610300494.4A patent/CN106023151B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104021566A (en) * | 2014-06-24 | 2014-09-03 | 天津大学 | GrabCut algorithm-based automatic segmentation method of tongue diagnosis image |
| CN104933723A (en) * | 2015-07-21 | 2015-09-23 | 闽江学院 | Tongue image segmentation method based on sparse representation |
Non-Patent Citations (4)
| Title |
|---|
| Tongue Color Analysis for Medical Application;Bob Zhang 等;《Evidence-Based Complementary and Alternative Medicine》;20130422;第1-11页 |
| 基于差分统计方法的舌象纹理特征的分析与识别;许家佗 等;《上海中医药大学学报》;20030930;第17卷(第3期);第55页 |
| 复杂环境下中医舌诊关键技术研究;陈碧黎;《道客巴巴——http://xueshu.baidu.com/s?wd=paperuri%3A%28b4f6e7d78daabb6f1aa68b3191b62a46%29&filter=sc_long_sign&tn=SE_xueshusource_2kduw22v&sc_vurl=http%3A%2F%2Fwww.doc88.com%2Fp-3197552581670.html&ie=utf-8&sc_us=3834659808226783687》;20150124;第3、18-41页，图1.2 |
| 自然环境下舌诊图像偏色检测及其颜色校正方法;刘齐 等;《厦门大学学报(自然科学版)》;20160331;第55卷(第2期);第278页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106023151A (en) | 2016-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106023151B (en) | Tongue object detection method under a kind of open environment | |
| Cerutti et al. | Understanding leaves in natural images–a model-based approach for tree species identification | |
| CN104834898B (en) | A kind of quality classification method of personage's photographs | |
| CN109523535B (en) | Pretreatment method of lesion image | |
| CN109635871B (en) | Capsule endoscope image classification method based on multi-feature fusion | |
| Ramakanth et al. | Approximate nearest neighbour field based optic disk detection | |
| CN111524080A (en) | Face skin feature identification method, terminal and computer equipment | |
| SG190730A1 (en) | Method and an apparatus for determining vein patterns from a colour image | |
| US20160314567A1 (en) | Systems and methods for image/video recoloring, color standardization, and multimedia analytics | |
| CN108388874A (en) | Prawn morphological parameters method for automatic measurement based on image recognition and cascade classifier | |
| Casanova et al. | Texture analysis using fractal descriptors estimated by the mutual interference of color channels | |
| Jadhav et al. | Soybean leaf disease detection and severity measurement using multiclass SVM and KNN classifier | |
| CN106650606A (en) | Matching and processing method for face image and face image model construction system | |
| Mathew et al. | Transform based bleeding detection technique for endoscopic images | |
| CN111062936B (en) | Quantitative index evaluation method for facial deformation diagnosis and treatment effect | |
| CN110458792A (en) | Method and device for evaluating quality of face image | |
| CN110021019A (en) | A kind of thickness distributional analysis method of the AI auxiliary hair of AGA clinical image | |
| Junod et al. | The development of an automatic recognition system for earmark and earprint comparisons | |
| Montenegro et al. | A comparative study of color spaces in skin-based face segmentation | |
| Nahid et al. | Local and global feature utilization for breast image classification by convolutional neural network | |
| Bogo et al. | Automated detection of new or evolving melanocytic lesions using a 3D body model | |
| Fathy et al. | Benchmarking of pre-processing methods employed in facial image analysis | |
| CN109509194B (en) | Front human body image segmentation method and device under complex background | |
| CN116386118A (en) | Drama matching cosmetic system and method based on human image recognition | |
| CN109753912A (en) | A kind of multi-light spectrum palm print matching process based on tensor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |